Project ideas from Hacker News discussions.

Desert Ant Labs: local, fast models that run on device

📝 Discussion Summary (Click to expand)

1. Model performance and technical details
Users repeatedly note that Voz is essentially an ANE‑optimized Parakeet v3 with custom inference, achieving extraordinary speed on Apple silicon.
- “It's an ANE optimized version of Parakeet, with our own inference, which enabled us to push performance to about 300x realtime speed on an iPhone 16/17. Our next gen Voz model is trained from scratch and will be at least twice as fast.” – pveugen
- “at first i got very excited about a new fast transcription model (voz) but turns out its just parakeet v3 with some new inference code which is macOS/iOS specific” – sipjca

2. Business model and licensing concerns
The free‑tier (≤100k MAU) and the rationale for ongoing payments spark debate over fairness and sustainability.
- “Every model is free up to 100k monthly active devices. No tokens, no logins…. I dunno about the business model though…. why should you continue getting money off me and my customers?” – 1dom
- “Because they own the IP and they get to decide the terms of how it’s licensed…. A 100K free device license is a generous gift to small companies.” – Aurornis
- “How is it not obvious and fair that they are asking you to pay them when you see success (defined as >100K MAU)?” – handfuloflight

3. Platform availability and cross‑platform desire
Many commenters want the models to run beyond iOS/macOS—on Android, web, Windows/Linux—and note the current iOS‑first focus as a limitation.
- “A lot of the models would be useful in a web context… but they look like most of them are iOS only, few have a node package… benchmarks on modern iPhones so I doubt it would be that fast on a 20$ VPS.” – ashenke
- “Just a few are iOS first (pure practical timing/sequencing). We plan to make all models available cross-platform in the coming weeks.” – pveugen
- “I wonder why they only support Apple platforms, citing CoreML. Doesn't Android have a similar framework, ML Kit?” – illright


🚀 Project Ideas

VozCross: Cross-Platform SDK for Desert Ant's Voz Transcription Model

Summary

  • Enables iOS, Android, WebAssembly, and desktop (Python/Node) usage of Desert Ant's Voz on-device transcription model.
  • Solves the lack of Android/web/VPS support highlighted by HN commenters who want fast transcription beyond iPhones.

Details

Key Value
Target Audience Mobile app developers, web SaaS builders, and indie hackers needing offline speech‑to‑text on any device
Core Feature Unified SDK exposing Voz’s 300x‑realtime transcription via platform‑specific bindings (Swift, Kotlin/Java, WASM, Python pip, Node npm) with CPU/GPU fallback
Tech Stack Swift (iOS), Kotlin/Android Studio, Rust/WASM, PyO3/Python Node addon, C++ core, CMake build
Difficulty Medium
Monetization Revenue-ready: Per‑app SDK license (tiered by monthly active devices)

Notes

  • Commenters said: “A lot of the models would be useful in a web context… they look like most of them are iOS only… I doubt it would be that fast on a $20 VPS” – ashken; and “I would love to use Voz and Ear… for platform availability – meaning macOS, Windows and Linux” – lukevp.
  • Provides the cross‑platform reach developers requested, opening server‑side and VPS use‑cases while preserving the on‑device privacy advantage.

DocuStruct: PDF‑to‑JSON Schema Extraction Using Desert Ant’s Schemer + OCR

Summary

  • Extracts text from PDFs (via built‑in OCR or external Tesseract) and feeds it to Desert Ant’s upcoming Schemer model to produce user‑defined JSON schemas.
  • Addresses the request for “something that can process pdf files into a json schema” and title/tag generation.

Details

Key Value
Target Audience Data engineers, content platforms, and developers needing structured output from unstructured PDFs
Core Feature Upload PDF → OCR text extraction → Schemer model → JSON output matching a supplied schema; includes CLI, REST API, and desktop GUI
Tech Stack Python (FastAPI), PyTorch/ONNX runtime for Schemer, Tesseract OCR, React/Electron for GUI, Docker for deployment
Difficulty Medium
Monetization Revenue-ready: Pay‑per‑document processed (credits) or monthly subscription for high volume

Notes

  • nullbio noted: “The most useful one for me would be something that can process pdf files into a json schema. Title and tag generation from a post would also be useful.” Capevace added that once image support arrives, Schemer could enhance tools like struktur.sh.
  • Delivers a practical pipeline turning PDFs into structured data without manual labeling, matching the community’s desire for OCR‑on‑steroids and text‑to‑JSON capabilities.

Read Later